Shoryaku AI's free CSV sales diagnosis

Shoryaku AI has published CSV sales diagnosis. Pick one CSV exported from an online shop or an ad management screen, and sales, average order value, the daily trend, and concentration by product come out on the spot.

Signup and login are both unnecessary. The file you select is not sent to a server. Reading and aggregation both finish inside the browser.

I am building Shoryaku AI ( https://shoryaku.net ), a marketing-support SaaS. It is a feature that can turn the state “I have an exported CSV, but I cannot find the time to open it” into numbers as it is. Here is what comes out, and how to use it.

Use is 3 steps

1. Open CSV sales diagnosis (there is no login screen)
2. Pick one CSV you have on hand
3. The aggregation result comes out

That is all. If you also pick last month’s CSV, the increase or decrease against the previous period is included.

Numbers it can produce

It looks at what is inside the CSV and automatically chooses what it can produce.

For a calculated metric, the formula is shown on the screen as it is (for average order value, “sales ÷ order count”). A number whose origin is unknown is not placed.

If you run the order sample distributed on the diagnosis page (12 orders), you get a sales total of ¥168,200, an average order value of ¥14,017, and a breakdown by product.

The reason it shows “what share of the whole the top 5 take” is that what you do next changes depending on whether sales lean on a part of the products. If one product accounts for half, the first priority is not to let that inventory and those ads run out.

Also, rates and unit prices are not summed. Click-through rate and CPA go wrong if you average the rate on each row, because the weight of the row disappears, so they are recalculated from the totals. Even if the export contains a “click-through rate” column, that column is only placed as a reference value.

Feature 1: The file is not sent

I think many people brace when they hear “upload a CSV.” This feature does not send the file anywhere.

Reading, removal of personal information, aggregation, and display all run inside the browser you are using. The file does not reach Shoryaku AI’s server, or anywhere else. The aggregation result is not saved either.

In place of saving, when you want to compare with the previous period, you pick 2 files at the same time (3 or more is not supported).

Feature 2: Names, addresses, and phone numbers are removed automatically

An order CSV contains a name, an address, a phone number, and an email address.

These columns are left out of the aggregation automatically at the moment they are read. The decision is not finished by a match on the column name alone. It looks at both the column name and the shape of the values. It is removed for a column named “email address,” and also for a sequence of values such as taro@example.com. A column that could not be judged is unused by default.

An identifier such as an order number or a customer ID is hashed inside your device before it is counted, because leaving it out would make the order count impossible. There is no mechanism for restoring the original value.

The removed columns are listed on the screen. I thought it is easier to try when you can see what is not being used.

Feature 3: A CSV can be read as exported

Export files have a different shape for each service. They are set up so that you can pick them as they are, without fixing them by hand.

• Character encoding supports UTF-8 (with BOM and without) and Shift_JIS
• The delimiter is judged automatically among comma, tab, and semicolon
• Even if a report-name row or a period row is at the start, it looks for the header row and skips ahead of it
• Rows that start with “合計”, “総計”, or “Total” are excluded automatically so they are not counted twice
• Up to 5MB and 50,000 rows

What is not supported is written here as well. For an Excel file (.xlsx), export it to CSV or TSV and then pick it. A table with the months across the columns (product name, January, February, …), a header in two rows, and merged cells are not supported.

Feature 4: The column judgment can be corrected on the spot

Which column is sales and which column is the date are judged automatically, but that judgment is not adopted in silence.

Before the aggregation, a confirmation screen appears, and “columns used for the judgment,” “columns only shown as a reference value,” and “unused columns” are displayed separately. If the judgment is off, changing the role on the spot reruns the aggregation.

You can try the samples first

It is fine if you do not have a CSV on hand. The diagnosis page has 4 kinds of samples: orders, ad results, keyword results, and access analytics (the contents are made-up, for explanation).

• The order sample contains name and address columns, so you can see those columns actually being removed
• The ad sample has a report name at the start and a total row at the end, so you can confirm that it can be read as it is
• The keyword sample is tab-separated and has no date column, so you can see that without a date the trend cannot be produced

AI has not been used up to this point

This screen does not call an AI even once. The total, the average order value, the trend, and the concentration are all fixed calculations. If an AI is asked to calculate a total of several thousand rows, the claim of “with a basis” stops holding.

In exchange for that, what comes out here stops at the numbers. “How to read those numbers, and what to do next” is the role of Shoryaku AI itself. Under the diagnosis result, it shows “things not yet known” from metrics that were actually absent from that CSV. If only sales are in the file, that item is “how much was spent to produce this result.” For ad results, a different item comes out.

In Shoryaku AI itself, you connect Google Ads, Meta Ads, Google Analytics, Search Console, and Instagram, and it produces prioritized moves together with the reason, the steps, and the numbers to watch. This is not a free plan. You can try it on a 14-day free trial (no card registration at the start).

If you want to look at the numbers first, go ahead with no registration.

→ Open CSV sales diagnosis (login is unnecessary, and the file is not sent)

• Product
• Marketing
• SaaS